Add error check to groupToEvents so we don't blindly add error values (#39404) (#39504)

* add error check to perfmon groupToEvents

* linter...

* tests

* add changelog

* fix changelog

* make linter happy

(cherry picked from commit 760208b)

Co-authored-by: Alex K <[email protected]>
Co-authored-by: fearful-symmetry <[email protected]>

Author: mergify[bot]

CHANGELOG.next.asciidoc

@@ -174,6 +174,7 @@ https://github.com/elastic/beats/compare/v8.8.1\...main[Check the HEAD diff]
 
 *Winlogbeat*
 
+- Fix error handling in perfmon metrics. {issue}38140[38140] {pull}39404[39404]
 
 *Elastic Logging Plugin*
 

metricbeat/module/windows/perfmon/data.go

@@ -20,6 +20,7 @@
 package perfmon
 
 import (
+	"errors"
 	"fmt"
 	"regexp"
 	"strconv"
@@ -48,7 +49,7 @@ func (re *Reader) groupToEvents(counters map[string][]pdh.CounterValue) []mb.Eve
 				// The counter has a negative value or the counter was successfully found, but the data returned is not valid.
 				// This error can occur if the counter value is less than the previous value. (Because counter values always increment, the counter value rolls over to zero when it reaches its maximum value.)
 				// This is not an error that stops the application from running successfully and a positive counter value should be retrieved in the later calls.
-				if val.Err.Error == pdh.PDH_CALC_NEGATIVE_VALUE || val.Err.Error == pdh.PDH_INVALID_DATA {
+				if errors.Is(val.Err.Error, pdh.PDH_CALC_NEGATIVE_VALUE) || errors.Is(val.Err.Error, pdh.PDH_INVALID_DATA) {
 					re.log.Debugw("Counter value retrieval returned",
 						"error", val.Err.Error, "cstatus", pdh.PdhErrno(val.Err.CStatus), logp.Namespace("perfmon"), "query", counterPath)
 					continue
@@ -69,7 +70,9 @@ func (re *Reader) groupToEvents(counters map[string][]pdh.CounterValue) []mb.Eve
 			if _, ok := eventMap[eventKey]; !ok {
 				eventMap[eventKey] = &mb.Event{
 					MetricSetFields: mapstr.M{},
-					Error:           fmt.Errorf("failed on query=%v: %w", counterPath, val.Err.Error),
+				}
+				if val.Err.Error != nil {
+					eventMap[eventKey].Error = fmt.Errorf("failed on query=%v: %w", counterPath, val.Err.Error)
 				}
 				if val.Instance != "" {
 					// will ignore instance index
@@ -93,9 +96,11 @@ func (re *Reader) groupToEvents(counters map[string][]pdh.CounterValue) []mb.Eve
 		}
 	}
 	// Write the values into the map.
-	var events []mb.Event
+	events := make([]mb.Event, len(eventMap))
+	iter := 0
 	for _, val := range eventMap {
-		events = append(events, *val)
+		events[iter] = *val
+		iter++
 	}
 	return events
 }
@@ -111,7 +116,7 @@ func (re *Reader) groupToSingleEvent(counters map[string][]pdh.CounterValue) mb.
 				// Some counters, such as rate counters, require two counter values in order to compute a displayable value. In this case we must call PdhCollectQueryData twice before calling PdhGetFormattedCounterValue.
 				// For more information, see Collecting Performance Data (https://docs.microsoft.com/en-us/windows/desktop/PerfCtrs/collecting-performance-data).
 				if val.Err.Error != nil {
-					if val.Err.Error == pdh.PDH_CALC_NEGATIVE_VALUE || val.Err.Error == pdh.PDH_INVALID_DATA {
+					if errors.Is(val.Err.Error, pdh.PDH_CALC_NEGATIVE_VALUE) || errors.Is(val.Err.Error, pdh.PDH_INVALID_DATA) {
 						re.log.Debugw("Counter value retrieval returned",
 							"error", val.Err.Error, "cstatus", pdh.PdhErrno(val.Err.CStatus), logp.Namespace("perfmon"), "query", counterPath)
 						continue

metricbeat/module/windows/perfmon/data_test.go

@@ -28,6 +28,67 @@ import (
 	"github.com/elastic/elastic-agent-libs/mapstr"
 )
 
+func TestGroupErrors(t *testing.T) {
+	reader := Reader{
+		config: Config{
+			GroupMeasurements: true,
+		},
+		query: pdh.Query{},
+		log:   nil,
+		counters: []PerfCounter{
+			{
+				QueryField:   "datagrams_sent_per_sec",
+				QueryName:    `\UDPv4\Datagrams Sent/sec`,
+				Format:       "float",
+				ObjectName:   "UDPv4",
+				ObjectField:  "object",
+				ChildQueries: []string{`\UDPv4\Datagrams Sent/sec`},
+			},
+			{
+				QueryField:    "%_processor_time",
+				QueryName:     `\Processor Information(_Total)\% Processor Time`,
+				Format:        "float",
+				ObjectName:    "Processor Information",
+				ObjectField:   "object",
+				InstanceName:  "_Total",
+				InstanceField: "instance",
+				ChildQueries:  []string{`\Processor Information(_Total)\% Processor Time`},
+			},
+			{
+				QueryField:    "current_disk_queue_length",
+				QueryName:     `\PhysicalDisk(_Total)\Current Disk Queue Length`,
+				Format:        "float",
+				ObjectName:    "PhysicalDisk",
+				ObjectField:   "object",
+				InstanceName:  "_Total",
+				InstanceField: "instance",
+				ChildQueries:  []string{`\PhysicalDisk(_Total)\Current Disk Queue Length`},
+			},
+		},
+	}
+
+	counters := map[string][]pdh.CounterValue{
+		`\UDPv4\Datagrams Sent/sec`: {
+			{Instance: "", Measurement: 23},
+		},
+		`\Processor Information(_Total)\% Processor Time`: {
+			{Instance: "_Total", Measurement: 11},
+		},
+		`\PhysicalDisk(_Total)\Current Disk Queue Length`: {
+			{Instance: "_Total", Measurement: 20},
+		},
+	}
+
+	events := reader.groupToEvents(counters)
+	assert.NotNil(t, events)
+	assert.Equal(t, 3, len(events))
+
+	for _, event := range events {
+		assert.NoError(t, event.Error)
+	}
+
+}
+
 func TestGroupToEvents(t *testing.T) {
 	reader := Reader{
 		config: Config{